Google DeepMind releases Gemma 4: Frontier multimodal intelligence on device

AI Impact Summary

Google DeepMind is releasing Gemma 4, a new family of multimodal models designed for on-device intelligence. These models, available on Hugging Face, offer significant improvements over previous versions, including support for image, text, and audio inputs, and a focus on efficient quantization for deployment on various devices. The models' out-of-the-box capabilities and impressive performance scores (1452 LMArena score for the 31B dense model) make them a compelling option for developers building on-device AI applications.